working paper
department
of economics
KfJ-.-."^---
-•'.,
-.-
INFLATION, UNEMPLOYMENT, AND PUBLIC
OPINION POLLS.
Stanley Fischer and John Huizlnga*
Number 256
April 1980
massachusetts
institute of
technology
i
50 memorial drive
cahfibridge,
mass. 02139
lUO
£:
,,,,:,
5
ES^-^'
INFLATION, UNEMPLOYMENT, AND PUBLIC
OPINION POLLS.
Stanley Fischer and John Huizinga*
Number 256
April 1980
* Department of Economics, M.I.T. Research support from the National Science
Foundation is acknowledged with thanks. We are grateful for assistance and/
or advice from Robert Burkhardt , Robert Comment, Thomas Juster, Michael
Kagay, Warren Mitofsky, Franco Modigliani, Lucas Papademos, Marilyn Potter
of the Roper Center at the University of Connecticut, and the Gallup
Organization.
Preliminary results of the research described in this paper
were presented in Fischer and Modigliani (1977).
1373335
Digitized by the Internet Archive
in
2011 with funding from
IVIIT
Libraries
http://www.archive.org/details/inflationunemplo256fisc
INFLATION, UNEMPLOYMENT, AND PUBLIC OPINION POLLS.
Stanley Fischer and John Huiztnga*
It
is well known that
public opinion polls report that the public
regards Inflation as a more serious problem than unemployment.
In this
paper we examine public opinion surveys with the aim of discovering what,
if anything,
the poll results me^n
,
We f^escribe in some detail the
questions about inflation and unemployment asked in the surveys
and
analyze the determinants, in both cross-section and time series, of
answers to those questions.
The tjrpical opinion poll question about inflation and unemployment
is "Which of the two problems
— inflation
or unemployment
— do
you think
will cause the more serious economic hardship for people during the next
year or so?"
Section
I
discusses the possible issues to which opinion poll
questions about inflation and unemployment might be relevant.
Section II tab-
ulates a variety of questions that have been asked, along with the responses,
and examines the views expressed and their consistency, both internal and
across different polls.
Section III presents a regression analysis of the
cross-sectional determinants of responses to the Michigan poll question
cited above, while Section IV studies determinants of changes over time
In aggregate responses to the same question.
Section V brings the analysis
of the preceding sections together in an attempt to answer the question of
what the poll results mean.
*Department of Economics, M.I.T. Research support from the National Science
Foundation is acknowledged wit?i tV.Tnks. We are grateful for assistance and/
or advice from Robert Burkhardt, Robert Comment, Thomas Juster, Michael
I^gay, Warren Mitofsky, Franco Modigliani, Lucas Papademos, Marilyn Potter
of the Roper Center at the University of Connecticut, and the Gallup
Organization.
Preliminary results of the research described in this paper
were presented in Fischer and Modigliani (1977).
-2-
I.
The Issues
.
Public opinion questions about inflation and unemployment have
addressed three basic issues:
(1)
the relative costs of inflation
and unemployment to society (2) the relative costs of inflation and
unemployment to the individual
(3)
whether economic policy should be
directed towards reducing inflation or reducing unemployment.
The last
issue of course presumes that there is such a trade-off.
Ideally, the evaluations of issues (1) to (3) would have some
explicit consideration of the time paths of inflation and unemployment
being considered.
For instance, individuals might be asked to rank the
costs of 1% more unemplojnnent for a year against those of an inflation
rate that is permanently higher by x%.
Unfortunately, the typical
question asks only about the costs of inflation and unemployment (perhaps
over the next year), without any explicit indication of the possible
tradeoffs between them.
Given this type of question, there can be a real distinction between
a person's views on issues (1) and (2) and his view on issue (3).
For
example, someone who believes society should be willing to pay the price
of a temporary recession to lower the inflation rate permanently might
nonetheless answer at the beginning of a recession that unemployment
will cause more serious hardship than inflation over the next year.
We will see below that questions relevant to all three issues
defined above
to
are
asked in the polls.
However, there is little attempt
make precise the tradeoff that the individual is being
asked to consider.
-3-
II.
Poll Questions and Answers
.
A large number of regularly conducted opinion polls ask questions
about inflation and unemployment.
2
In this section we report selectively
on poll questions that relate to specific issues described beloT/.
1
.
T he Overall Importance of the Inflation and Unemployment Problems
The first question in Table
organization since 1935.
1
.
has been asked by the Gallup
The question is of course open ended and there-
fore makes it possible to ascertain the importance attached to inflation
and unemployment relative to other problems facing the nation.
Given that
the poll started during the Great Depression and continued through several
wars,
changes in the relative ranking of economic problems should not be
attributed solely to
Figure
since 1939.
3
1
increasing awareness of
their importance.
shows the history of responses to the Gallup poll question
With only two exceptions, over 5Q% of the respondents
named either inflation or unemployment as the most serious problem
facing the country in every poll since September 1973; on several occasions
the combined percentages were above 80%.
Also noticeable is how quickly after
World War II concern over unemployment faded; thereafter it was essentially
only in recessions that unemplo3rment was ranked as a more serious national
problem than inflation.
Given the prominence of the inflation and
un'^-
employment issues since 1973, it is likely that the public by now has well
formed opinions on their relative importance.
It is perhaps not necessary to point out that Figure 1 shows the
public generally regards inflation as a more serious problem than unem-
ployment-^as is confirmed also by Figure 3 below.
TABLE
1
Opinion Survey Questions on Inflation and Unemployment
Organization
American
Institute of
Public Opinion
What do you think is the most
important problem facing this
country (sometimes, this section
of che country) today?
G
1974-Present
Which of the two problemsinflation or unemployment - do you
think will cause the more serious
economic hardship for people
during the next year or so?
M
1976-Present
Which do you think is the most
important problem facing the
country today - unemployment
N
Mici;i3an
Nev York Times/
Code*
1935-Present
(Gallup)
Institute for
Social Research:
University of
Question
Period
PT»P
or inflation?
The Harris
Poll
Poll
Sources
•-•
.
:
1974-Present
had to choose, which do
If you
you think is a more serious
problem for the country today rising prices or high
unemployment ?
HI
1975-Present
If you had to choose, which do
you think is a more serious
problem for you and your family
today - rising prices or high
unemployment?
H2
Gallup:
The Gallup Opinion Index November 1978, Report #160; Harris:
Press Release, March 20, 1978, ISSN 0046-6875; Michigan: Mimeos
released by Survey Research Center; New York Times/CBS: Telephone
conversations and computer print-out. *Code is for coding in Table 2.
,
o
H
O
C
o
l-t
a
o
o
a
o
3
3
a
o
<:
I
i-i
o
I
00
c
-1
3
I
(!)
ON
o
.
-4-
2.
Consistency of Responses Across Polls
.
1
ask for a direct comparison of
the costs of inflation and unemployment.
Although, as noted above, the
The remaining questions in Table
comparison is not well defined, it is worthwhile checking for consistency
of responses in different polls.
Table
(N)
The
2
shows the results of the Michigan (M)
,
New York Times/CBS
and Harris (Hi) polls to similar questions asked around September 1976.
aggregate responses are similar.
Results of the Gallup poll taken
at the same time are out of line with those of the other three polls.
However, given the possibility of answers other than inflation or unemployment to the Gallup question,
of
it
would be necessary to assume the irrelevance
independent alternatives for the Gallup results in Table
2
to be expected to
coincide with those of the other polls.
The Gallup poll results are even more different in February 1975.
Indeed, there does not appear to be a systematic relationship between
answers to the Gallup and other polls.
This likely means that
alternatives excluded when individuals are asked to rank inflation and
unemployment affect the comparison.
Although the differences between the Michigan and Harris polls are
larger in February 1975 than in September 1976, both polls show unemployment being regarded as the more serious problem in the earlier period.
Further, the difference between those poll responses is at its largest
in early 1975, a period of rapidly changing unemplo3^ent
The basic picture is thus one of substantial similarity of responses
to similarly worded questions in different polls.
.
.
TABLE
2
Per cent Responding
^
Quest ion
G
Period
1976
Inflation
Unemployment
60.3
39.7
Oct.
M
1976
Aug/ Sept
55.1
44.9
N
1976
52.2
47.8
1976
Aug
51.3
48.7
1975
75.0
25.0
1975
32.8
67.2
Sept
HI
M
Feb.
HI
1975
Feb
42.0
58.0
H2
1975
84.1
15.9
Notes: Those who answered Don't Know or Both are dropped from these samples.
-5-
3
.
Social versus Personal Costs of Inflation
The different responses, shown In Table
In
Table
1
CHl and H2)
.
2,
to the last two questions
reveal the clear distinction individuals draw
between the iadividual and social costs of inflation and unemployment.
In February 1975 over 8Q% of the respondents to the Harris poll said that
rising prices were a more serious problem for their families than was
unemployment.
But at the same time less than half the respondents thought
inflation was a more serious problem for society.
It
is thus clear that individuals' responses to the standard
inflation/unemployment question, M in Table 1, do not reflect only
perceptions of their self-interest.
In Section III of the paper we
investigate the factors affecting individual perceptions of the social
costs of inflation and unemployment.
4.
Is There a Trade-Off Between Inflation and Unemployment ?
Table
3
presents a series of questions dealing with the trade-off
between inflation and unemployment.
off exists.
The first asks whether such a trade-
A large majority of the respondents believed
it does.
The
question is not sufficiently detailed to make it clear whether individuals
believe there is a long-run trade-off.
5.
Should Unemplo3nnent be Increased to Reduce Inflation ?
The remaining questions in Table
3
relate to whether the Inflation
unemployment trade-off should be used to reduce inflation.
In all
the polls shown, the responses are that policy should be directed towards
reducing unemployment.
This applies to polls from October 1975 through
April 1979, even when the Michigan poll for the same period (the standard
question)
shows more people regarding inflation as being a more serious
1
'
^
7}
-o
« o
<U •H
CrJ
M
r>
4-1
r-t
r-4
O
CT.
PQ
;
(U
rH PM
iH
o
PM
C
d)
1
§
en
O
M
D.
CO
OO
•H
O
^a m
-3-
CTi
1^
CM
CO
CO
r>
VD
00
cs
<!•
u-1
isS
C3
1
c
o
t-l
U-i
-H
C
4-1
M
CO
(U
4J
*
)-l
O 3
Z
CO
T)
1
1
0)
(U
U
O 3
VO
to
r-4
S
o
m
M
O 3
4J
4-1
:^
r-l
w
d
3
c
O
^
UH -H
d
M
rH
^
4-10
O
<r
4-1
W
CO
"O
«
4J
3
to
O
CM
M ^
0)
c
o
o
o
0)
CO
w
CNi
<r
in
o
^-|
>-i
tu
0)
4-1
rH
CN
rH
d
O
<4H
-H
•-i
Md
4J
cn
1
to
,p
>
^-'
<u
to
to
CO
•<t
D
4J
m
u
o
CM
to
M
5
-a-
1
rH
u
C
CO
q
O U *H
CD
>
•
to
T-t
(1)
S
0)
6
V4
to
•H
^
^
CU
,H
iH 4J Xi
B
o
0)
•H
*j
4J
0)
m
3
O
(U
^
3
r-l
r^
•H
(0
n
.
^
C
•tH
14H
OJ
to
^O
CO
O
C
O
-H
^
to
4-1
>
CO
CO
0)
1,
S
O
0)
M-l
&
4-1
J=
£50
•H
,c
«
0)
0)
^1
C
0)
•
4-1
CO
-C
to
0)
-H
O
•H
>
O
o
a
0)
^
4-1
X: -H
4-1
.-1
P.
00
c
CX^
P.
3
to
6
3
c
3
>H
T-t
<u
:s
x:
00
-rH
W ^
rH rH
CO rH
S
Vj
0)
O
^o
a
0)
>i
•H
CO
e e
60
o
C
O
O
*H
a.
3
4-1
.4-11^-1
> 4-1
C 0)
o 4-1 4-1 O CO
T-^
O CO CO
>^
to
u
•H
T3
rH
i*-i
0)
^
4J
3
o
x;
60
^
O
13
)-l
14-1
0)
4-1
4-1
4-1
4-.
o-
-r-l
0)
CO
4-1
6 u
0)
flj
<D
O
t
u
(U
to
to
,p
s
S
x;
O
4J
e
4-1
rH
d
(U
o)
<4H
tu
d
3
M
vO to
r^ 4J
OA C
r^ tu
-
:s
W
•>
0)
N
01
(!
4.1
;
c:-.
"ri
C'
I—
4J
to
3
6C
3
-3
-a
CO
C
fw
CO
OJ
•H
s
l-l
-3
CO
u
r- C
r^ 0)
CM -O
V
.
CA
r^
a.
H <
^
d
o
t.
> w
o a
E- H
C
o
•
fc-
CD
<U
^.
-O
•>-d-
>
to 2
o
:s sr
en
>- r-.
m
>- rH
1
(U
i.
> o
O rH
o.
o. e
(U
^4
d
o 3
d 60
o d
•H -H
4-1
O
CO
3
l-l
^
•S
-o
60
o d
4J
(U
M
O
^
H
•H
3
J^ -H rH "n
(U
4-1 "O U-l
<4H
rH
Vu
4-1
to
q
»-i
4-1
4J
2i
4-1
4-1
4-1
>A
4-1
^
4-1
>-.
c d
E -H S
O ^4 O
d cu J-1
o
o d -o
4-1
4-1
4-1
60 4J
CO
4-1
(-1
4-1
c^-
•
•H -a
-MM
>^ O
)-i
4-1
•
•rl
CO
00
C
<D •ri
^ cq U-l •o<u
[5
e d
tH CO
o o
3
^
OJ -5
CJ 4-1
d
3 C <D 43
to
01
J=.
V
d rH ^ J •H
OO
0)
to
4-1
3 d (DO)
^4 -H
(U
o to E OJ O d d
to
tH o
•c ^
(D
>,
4-1
Q) X
<D
4-1
(0
^ o
e o
rH >, to ^J
,Q O ^ ^
<-{
H
ex rH CO 3 to
3 tU o -S
O -H B r-^ J=' O d
01
CO
CO
<4H
^O <D 4-1to S COf^
d d P^ w <u
to
to -H
CO
4-1
3 o 4-1 o e
to -H
tu
CO
CO
C
0,
rH ^ (U to rC d ^
^
tu
00
Q)
rH &
O ^ •"
CO
O >% P^ 0) >. > •H &-H O
0) X:
M -ri lu
u a>
O ^
•H
•H rH
•H
a
3 a
CO
OO
&. O
Vj
d -a o
0)
o e
o o
d
e •H 0) •" •rl -H S ^-^^ oj
OJ
o M C S^ ^1
d o to -o >
rJ t« a 3 c» CI. CO •H O B --^ Q)
;-.
o
d
o
<u
x;
1
4-1
14-1
u
o
3
o O
v^ 4-1 S
J3
OJ
to
"
0)
to
d
H-l
^
c
a
>-i
^5
iH
^-^5
o«
M
4-1
O
C
C
o
(0
.
CO
4-1
(U
0)
4-1
H
•
0)
^
CO
60
-H
to
-H
-H
(8
CO
.
J-l
e
>,
S
>^
iH
iH
l-i
•
O
O
c
o
a)
0)
l-l
CO
^
M
rH
v^ S
c
^
a
o
3 X
O O
0)
M-l
^0)
d
4-1
(U
4-1
.-1
•
Q)
c
CO
^ CO ,cOO taO
C
3
o c
O <U
o
P^ 4-1 >> u
13 •H
•H
•
1
Q)
to
14H
dj
1
)-i
<r
>-5
to
to
jn
•H
J3
d
CO
to
U ^
3 O
O -rn
n
4J
00
3
TJ
(0
<U
•H
Pi
o
U
f-i
Ph
o
(a
<U
o
<u
C
CO
r
o
00
CO
CN
CM
CO
in
CO
ro
CM
(Q
60
•H
U
O
^
B
rH
O
u
C7^
«3-
C O
o C
CSJ
o
CN
CO
CM
in
3
CO
0)
PS
c
o
o
CO
CO
u
-H
01
CO
iJ Ph
Pi
cr«
ro
CM
CO
O
m
CN
Ed
5
00
(1)
S5
>.
CO
CO
tH
CO
4J
V4
(U
0)
CO
•H
a.
CO
(U
M o
4J
4-1
U
J3
0)
o
P^
CO
CO
60
0)
CO
O
O
c
o
CJ
o O
C
0)
M ^
o
6
O
3
M
> O >
-H
M-l
-5
CO
CO
- J=
"
> e c
CO
<u
o
C "H
3 4J
CO
& x; .-I
bO
QJ
J=
&0
P<
C
•u
w
4J
o S
CO
<u
3
ca
CO
0)
!>•
-O T3
U C
(U
1
1
u
o
CO
•--*
^
(-1
t-ct
a.
OJ
CO
?^
CD
CO
,H
a.
O
c
e
^ •HMO)
•H
-H
u
0)
jj
C
C3
O
as
0)
u
CO
en
0)
c^
in
CO
60
P
O
CU
<U
Di
l-i
(U
d)
(X
p.
o
pc:
,§
Du
O
c o
00
u
a
o
o
o
0)
r-
0)
O
u
n
3
o
CO
CO
CTi
o
o
4-1
4-1
o
o
CO
-6-
problem than unemployment.
There is no necessary inconsistency between believing that inflation
imposes greater costs on society than unemployment, and not being willing
to incur the costs of higher unemployment to fight inflation.
In terms
of Figure 2, assume that public has indifference curves between inflation
and unemployment.
the southwest.
These are negatively sloped, with utility rising towards
They are drawn linear, in the absence of a priori
restrictions on their concavity.
Presumably individuals answer the
question of whether inflation or unemployment is more costly by considering
the slope of the Indifference curves.
Figure
2
also contains a Phillips curve, labelled PP, showing the
terms at which individuals believe inflation and unemployment can be
traded off.
Now consider an individual who expects to be at point A
in the next period.
If the slope of the indifference curve at A is
sufficiently flat, he may well say that inflation is the more serious
problem.
Yet if PP is even flatter, as illustrated, he will want policy
to be directed at reducing unemployment.
There is an alternative interpretation of the meaning of the answer
to the standard poll question that makes it possible for individuals to
answer that inflation is the more serious problem but that policy
should be directed towards reducing unemployment.
Individuals might be
considering whether society would be better off if it could eliminate
inflation entirely while keeping the unemployment rate at its expected
level, or by eliminating unemployment (perhaps reducing it to the natural
rate) while keeping inflation at its expected level.
In terms of Figure 2,
Inf lat: on
Unemployment
Figure
2.
-7-
they would answer that inflation is the more serious problem
Indifference curve through point U
if the
were below that through point
11
.
The policy choice would still be determined by the relative slopes for the
PP ?nd Indifference curves througn point A.
6.
8
Links Between Inflation and Nominal Income
Table
4
.
contains the last set of survey questions and responses
in this section.
They are designed to shed some light on what we
call the "misunderstanding hypothesis", which seems to have developed in
response to the difficulty of explaining the economic costs of inflation
relative to those of unemployment.
The hypothesis claims that the
public dislikes inflation because it fails to link increases in its own
nominal income with inflation, on which, however, increases in
the cost
of living are blamed.''
The evidence for the misunderstanding hypothesis is difficult to find.
It may originate from the 60's when
it
inflation was sufficiently low that
may not have been systematically incorporated in individuals'
thinking.
The evidence in Table 4 is mixed.
is implausible that 70% of the respondents
It
to the Roper poll in April
1978 had their incomes rise by less than the cost of living over the
previous few years.
The answers to the December 1978 NBC/Associated Press
poll, however, show
little support for the misunderstanding hypothesis.
Part of the difference between
4
the results of the two first polls in Table
may result from the difference between "family" and "your" income,
since
rising family income in the seventies has been in large part attri-
butable
to increasing labor force participation.
c o
0)
Si
3
3
0)
M
3
fO
C/3
O
O
4-1
13
4J
o
4J
P.
2;
(U
<U
A!
O
O
C
O
<U
to
CO
M ^
O
(0
P<
0)
td
>l 4J
o
3
O
O
CM
3
en
NO
0)
CO
(0
tt-l
(1)
O
60
en
>
o
-H
^J
(U
O
-*
CN
6
O
&
CO
CO
Q)
M
00
^
CO
CM
iH
CO
0)
OQ
a)
o
R
in
a
O
3
00
CO
a
0)
M
3
o o
C
iH
P.
to
to
0)
CJ
0)
?^
p.
•H 4J
to
^4
o
S
c o
p.
CO
c
u
:3
^
3 0) O
O > >>
^
to
<u
4-1
3 e
o
c & p<
CO
3
C3
o-
rP
O
CO
>
p>>
C
0)
^
(0
-H
CO
-H
CO
r-(
^
CO
OJ
0)
m
(1)
u
CO
CO
CO
CO
-H
CO
CO
r:
CO
rH
3
0)
4-J
>4-l
•H
to
o o
13
u
g 2
CO
4-1
a,
3^
O
v^
(u
4-1
d
O
60
CO
o
C 3
o o
1-1
•H
>^
iH
4-1
to
-H
4J
01
CO
u u
O CO
B
^-N
CO
CO
0)
(U
)-i
f^
3
o
>1
CO
Q)
OJ
13
IJ-I
CJ
t::
3
CU
O
60
(U
0)
U
4-1
CD
4J
e
O
o
&
CO
<D
CO
•H 13
>
01
CO
•H
^1
)-l
1-1
CO
<U
cw
60
CO
O
a:
rH
.i
CD
Q)
H
CO
•H
4J
•
^
•
0)
15
.5
4-1
^
OJ
O CO
e
3 o
O
o
O 4J
4-1
>,-H
^-N
CO
J2 42 rH
13
C
o 3 O
o O o
^u uCO a;to
o & 0)
4J
^
3 4-1
13 o
CO
>% n
J=
o
13
3 i-f
O 3
1
^i
60
c
<u
.H
>
<V
CO
1-t
r-{
O
4J
P.
p.
CO
I
3
3
CO
^4
OJ
<u
CO
rH
CO
J2
p-
4-1
CO
i-H
l+H
nQ
CO
3
3
3
13
CO
0)
JZ
3
4-1
(U
0)
CO
-H •H
0)
.H
CO
CO
CO
CO
4-1
:s
4J
-H
>h
><:
Jh
0)
>>
IS
>% p.
CO
Xl
C--
iH
to
0)
01
CO
t-l
•H 13
CD
4J
14-1
.c
60
c
S
>s,
4-1
4-1
•
o
c
0)
CI
CO
CO
0)
CO
M
o
P,
c
O S 3
0)
CU -H
u to
p. IM
o
CO
a
4J (U 3 •rH
CO
u
o
0)
•T3
iH iH CO
-H 3 -O
T-l
u
4-1
•
I
1+-I
c
co
4-1
^
CO
to
O
3
0)
CO
0)
CO
rH -
X!
I
M to 0)
in 4-1 o B U C
3 CO O
5 3
O 0) 60
o u >^
* •a •H
CO
M-l
O
CO
O P.
O ^
3 &
CO
o rH
Q)
O cd to 0) (U
60
c -H e
^
O P< O
•
to
CO
4-1
3 o
to
0)
M O C
60 o 0)
O "H
•H e P.
5
4-1
U
o
o C to 3
3 to O
6 tn -5 O P. >i
CO
<U
MOM
3
C
4-1
13
CO
0)
O
CO
CO
to
4-1
60
4-1
3
J2
*J
CO
(U
)-l
x:
u
I
o
o
o
<u
.3
CJ
W
?^
•
r-s ^H rH
<U
CO
.H
P.
CO
3
O
4-1
0)
4J
3
0)
a
C
O
iH
.H
•H
o
4J
PLi
t»
13
«r>.
0)
4-1
<yi
U
o
—(
p- >,
0)
CO
43*2
C
CO
Cl)
60 4J
u
O Q
to
o
o
CD
CO
•
T-',
rj
c
O
o
tu
O
•
-a
to
CO
X d
O o
0)
-H
^4
P.
l-i
a
<
4-1
to
i-H
o
r-.
CTi
.-1
4J
U
Dd
CO
*J
CO
**«N»,
00
P.
P.
<l
p<
to
OJ
u
<
-^
U
0)
O
»
CO
CO
Q)
PQ M
2: P^
O
ey»
i~»
a\
IH
<U
13
r-l
3
O
P.
iH
§
a
CO
dJ
l-i
o
o
VD
f-H
•H
M
P.
<
•>>
S
W
>-
01
o
u
3
o
to
-8-
The last question in Table
4
is included to show that when the
question is put directly, the public does see a connection between pay
raises and inflation.
But this is only weak evidence against the
misunderstanding hypothesis, whose proponents would not necessarily
question the ability of the public to understand that higher pay
raises will lead to price rises.
The misunderstanding is rather that
the public is supposed not to understand that inflation also raises
nominal incomes.
Although the evidence on the misunderstanding hypothesis is weak,
it
is significant that we have been able to find no strong evidence in
its support.
7.
Sensitivity to the Wording of the Questions
.
There is evidence that answers to poll questions are highly
sensitive to the wording of the questions.
Such difficulties do
not arise in the responses to the Michigan, Harris (Hi) and New York Times/
CBS poll questions in Table
1,
even though there are slight differences
in the wording of the questions.
One difference in the wording of the questions merits special
attention.
The Michigan and New York Times/CBS
polls ask about
inflation while the Harris poll asks about rising prices. It is
sometimes argued that there is a popular confusion between inflation
and "high prices".
Indeed, examination of the questions asked by the
poll takers suggests that they sometimes confuse high and rising prices.
For example, in October 1970 Gallup asked the following question:
—3
-9-
"When people around here go to vote on November 3rd for a candidate
for Congress, how important will inflation
be in their thinking..."
— the
high cost of living
In addition, Gallup clearly refers to the
high cost of living and inflation interchangeably in its own reports.
Despite this confusion by some poll takers, the evidence is that
associating inflation with high rather than rising prices does not
seriously affect the poll results.
As Table
2
shows, the Harris and
New York Times/CBS polls elicit essentially the same responses even though
the former asks about rising prices and the latter about inflation.
Of course, this does not necessarily mean that the public draws a
tight distinction between high and rising prices; it means only that
even when the distinction is made explicit
8.
poll responses are similar.
Summary .
The material reviewed in this section relates to the general issues dis-
cussed in Section
Section II.
1,
I
in several ways.
First, the results of the Gallup poll in
and those of the polls examined in Section II.
2
and II.
show that the public does indeed regard inflation as having a substantial
social cost relative to unemployment
— and
also that inflation and unemploy-
ment have in the seventies been regarded as the most serious of all
problems facing the nation.
Second, the Harris poll results discussed in
Section II. 3 show that the public draws a clear distinction between the per-
sonal and social costs of inflation and unemployment: the social costs of
inflation relative to unemployment are regarded as relatively lower than the
personal costs.
-
On..the third issue specified in Section I,
the public
-10-
generally
believes that economic policy should not aim to reduce inflation
increasing unemployment.
by
We have, in this section, attempted to
reconcile the apparent inconsistency between the belief that inflation
is the more serious problem than unemployment with the argument that
unemplojment should not be increased in order to fight inflation.
We turn now in Section III to an intensive examination ol the factors
determining individual responses to the standard (Michigan) poll
question.
III.
Determinants of Individual Responses
.
In this and the next section we examine in detail the determinants of
answers to the Michigan (standard) question listed in Table
I.
Because
the Michigan poll has been taken more regularly and over a longer period
than other polls that ask for a comparison of the costs of inflation and
unemployment, it provides sufficient data to do both cross-section and
time series analysis.
12
We start in this section with the cross section
analysis.
The Michigan question asks whether inflation or unemployment will
cause more serious economic hardship "for people" over the next year.
We assume that respondents answer this question on the basis of their
perceptions of social welfare.
Our implicit model of the individual is
that some set of charcteristics, which change slowly or not at all,
determines his underlying attitudes to inflation and unemployment as
summarized, say, in the indifference curves in Figure
2.
Given these
attitudes, current and/or expected rates of inflation then determine
-li-
the individual's actual rate of substitution between inflation and un-
employment and determine the response to the poll questions.
We have studied cross sectional data from five Michigan surveys,
those of Winter (February) 1975, Fall (November) 1975, December 1977,
Simple relationships between the answers
February 1978, and May 1978.
to the poll question and some characteristics of the respondents are
presented in Table
5,
for three of the surveys:
1975, and December 1977.
13
Winter and Fall
The Winter 1975 poll is one of the few in
which there is a clear majority who regard unemployment as the more
serious problem.
It
was taken at a time when the unemployment rate had
risen from 5.5% in August 197A to 7.9% in January 1975 and was about
to rise by another 1% in the next four months.
Prices (CPI) had risen
nearly 12% over those of a year before; the inflation rate was about
to fall to under 5% in first quarter of 1975 (seasonally adjusted CPI,
April over January, 1975).
The Fall 1975 poll was taken at a time when
both the unemployment and inflation rates were falling.
The December
1977 poll was taken at a time of falling unemployment and rising inflation.
It is useful in the cross section framework to categorize explanatory
variables into three classes.
or
First, some variables such as age, race
ethnic grouping, and sex are truly exogenous for the individual.
Second, characteristics such as income and education are unlikely to
te significantly influenced by attitudes toward inflation and unemployment,
even though these are in some sense choice variables for the individual.
^12-
Finally, variables such as home ownership, the extent of nominal
indebtedness, and political affiliation, would likely reflect under-
lying attitudes to inflation and unemployment.
Correlations between
this third type of variable and poll responses therefore have to be
The entries in Table
interpreted with care.
categories,
14
5
show, for a variety of
the percentage of those answering either inflation or
unemplojonent who answered inflation to the standard question.
For
simplicity we refer to this number as an index of inflation aversion,
or merely inflation aversion.
We start in Table
inflation aversion.
5
with the relationship between income and
It has been observed by Hibbs
(1976) that infla-
tion aversion is positively associated with income in the cross-section.
The overall relationship seen in Table
5
is one in which income and
inflation aversion are positively correlated, but it is only in the
Fall poll that the relationship is monotonic.
The second set of entries in Table 5 relate to employment status.
Here there is fairly high level of concern among the unemployed themselves about inflation.
This is another piece of evidence that
perceptions of individual and social welfare differ.
For
<-he
next two characteristics tabulated, the endogeneity issue
becomes important.
Home owners are on average more inflation averse than
those who do not own homes.
While this may appear surprising in light
of the fact that those who own homes have a good hedge against inflation,
it will be recalled that people are not answering in their personal
Interest.
Further, causation could run from inflation aversion
Table
5
Cross-Sectional
;
Responses to Michigan Surveys, by
Characteristics of Respondents.
Characteristic
Poll Date
by Category
Percentage Responding Inflation
0-4.9
5-9,9
10-14.9
15-19.9
20+
27.2
(114)^
33.9
40.5
29.9
(186)
(220)
(142)
39.2
(163)
50.8
52.7
53.3
59.8
(257)
(278)
(288)
(206)
61.5
(279)
43.6
59.4
(245)
63.4
(264)
62.9
63.5
(118)
(232)
(349)
Not in
Labor
Working >
20 hrs/wk
Working <
20 hrs/wk
Temporarily Unemployed
Laid Off
or looking
for work
(1)
Household
Income
$l,000's
75:1
75: IV
77:12
Force
(2)
Employment
Status
—
75:1
75: IV
77:12
Unavailabl e for this perdLod
54.1
(555)
56.2
41.7
66.7
50.8
(783)
(66)
(12)
(61)
57.1
(242)
61.6
66.3
28.6
54.2
(864)
(43)
(14)
(36)
Yes
(3)
Home
Ownership
—
—
75:1
No
Unavailable for this period
—
75: IV
56.3
(791)
51.4
(552)
77:12
59.9
56.8
(190)
(609)
Strong DemoDemoc. crat
Indep.
Democ.
Indep.
Indep.
Repub.
Repub.
Strong
Repub.
28.8
(146)
40.2
(107)
38.8
36.0
(75)
39.1
(124)
31.1
(170)
58.4
55.7
(118)
60.8
64.4
(196)
(91)
(A)
Politics
75.1
75: IV
77:12
Note:
32.8
(204)
46.5
51.3
(231)
(341)
—
53.1
(181)
(298)
Unavailable for this period
—
(61)
i^ntries in each cell are percentages responding inflation to the inflation/unemployment question, assigning "Both" equally to the inflation and unemployment categories, and
omitting "Don't Know", etc.
Numbers in parentheses indicate number of respondents with the relevant characteristic
who answered the inflation/unemployment question (excluding "Don't Know").
j
-13-
to home ownership:
home owners.
fear of inflation may cause individuals to become
In addition,
homeowners presumably have higher incomes than
non-owners and the ceteris peribus relationship may not be as shown
in the table.
The fourth entry in Table
shows a fairly strong relationship between
5
political affiliation and attitudes toward inflation, stronger in the
1975: IV poll than in the 1975:1 poll, however.
Both polls demonstrate that
among Democrats, the extent of political identification with the party
is strongly and positively correlated with an aversion to unemployment.
For Republicans there is the anomaly that strong Republicans worried
relatively more about unemployment than did other Republicans in 1975:1.
Nevertheless, it is appropriate to say that on average, inflation a-^arsion is
related to political affiliation.
As in the case of home ownership though,
the causation is not obvious.
The relationships in Table
5
obviously cannot be regarded as representing
the partial effects of the specified characteristics on attitudes tov/ard
Inflation and unemployment.
Accordingly, we have run a series of maximum
likelihood regressions in an attempt to isolate the effects of factors such
as income, education, age, sex, race, expected inflation and expected unemployment.
Political affiliation, home ownership, and employment status have
been excluded from the regressions.
for most periods.
We do not have the unemployment variable
The other two variables are excluded because of the
endogeneity problems discussed earlier.
The regression model, which assumes that the dependent variable can
take on one of three discrete values, is described in Appendix
3.
It is
-14'
assumed that the probability a person answers inflation is a function of the
slope of his indifference curve and that the slope is a function of the
regressors used in the equation.
The three choices for the dependent variable
are "Inflation", "unemployment", or "both".
The regression maximizes the
likeli-hood function by assigning weights to the regressors and choosing
an interval in which the individual responds "both" when asked which is
a more serious problem.
Regression results for all five periods we have examined are presented in
Tables 6 and
7.
Positive coefficients in Table
6
indicate that the probability
o^-^nswering inflation increases with, the right hand side variable^.
whether a respondent is black or hispanic (one category) or not.
on the va±ue one for whites and two for blacks and hispanics.
is one for males and two for females.
B /H indicates
The variable takes
The sex variable
We do not report the coefficients
defining the interval of indifference in which individuals answer "both", or
the constant terms, in order to reduce clutter.
Appendix
4
contains those
numbers, as well as a description of the scaling of the other right hand
side variables.
The three variables that enter most significantly are the two expectational
variables and B/H.
The coefficient on expected inflation is significantly
different from zero in every time period and those of B/H and expected
unemployment are significant in four out of five periods.
Clearly, individuals'
ranking of the costs of inflation and unemployment depends on the levels of
those variables.
The significant sign on the B/H variable indicates that
blacks and hispanics feel unemployment is a more serious problem for the
Table
6
Multinomial Probit Regression? Explaining Inflation/Unemployment Responses
;
Regression
Date of
Poll
-Log likelihood function
Independent Var•iables
Inceme
Education
.121(2.2) -.242(1.6)
.026(0.2)
.:oooco:o)
.022(0.9)
.050(0.8) -.171(1.1)
.093(0.8) -.008(0.
715.89
.010(0.4)
.055(0.8) -.062(0.4)
.240(1.9) -.027(1.8)
740.95
.025(1.1)
.156(2.4) -.264(1.6)
.245(2.0) -.025(1.7)
586.26
.027(0.9)
.164(2.2) -.320(1.7)
Zee.
vr
2
1119.49
.018(0.9)
Dec.
77
3
829.98
Feb.
78
4
May
7?.
5
-
1
(cont'd)
Expected
Unemployment
-.133(0.8) -.158(1.7) -.049(1.5)
2
(cont'd)
-.363(3.1) -.057(0.8) -.119(5.1)
.088(4.1)
3 (cont'd)
-.640(4.7) -.098(1.2)
114(3. Q)
.124(5.5)
4 (cont'd)
-.610(4.3) -.034(0.4) -.132(4.5)
.065(2.7)
5 (cont'd)
-.742(5.7) -.093(1.1) -.078(2.6)
.052(2.3)
B/ti
Sex
-.
-statistics in parentheses
Education
is a dummy variable for 16+ years of education.
t
-(-
2
.014(0.7)
1
1.
Age
-.106(0.7)
75
2.
.
Age
Education^-
Feb.
Notes:
,
Expected
Inflation
.081(3.2)
f^)
•
-15-
country relative to inflation than do whites, even after standardizing for
expectations and other variables, such as income, in the regressions.
None of the remaining demographic variables enter in a consistently
significant way, however.
The coefficient on income is positive in
each case, but with a t-statistic that is usually around unity.
Income per se is not strongly related to how the public views the
relative costs of inflation and unemployment; the simple relationship
found in Table
5
is misleading.
Next to B/H, the education variable
comes the closest to being consistently significant, with a positive
effect on inflation aversion that is significant in three periods.
Inflation aversion appears to increase with education until the post
college level, although the educational dummy for 16+ years of education
is never significant.
Age and age squared show some signs of contributing
explanatory power in later periods.
The sex variable has a negative
coefficient in each period, implying women are more worried about
unemployment than men, but the effect is never significant.
Table
7
presents calculations that help interpret the magnitudes
of the coefficients in Table
6.
The data in Table
7
are estimates of the
probability that individuals with the specified characteristics will
respond inflation to the poll question.
The standard case is that of
white male with twelve years of education,
income of $15-20,000, aged
35-44, expecting 5% inflation and no change in the unemployment
rate.
The standard case remains constant across time.
In Table 7
the characteristics are varied one at a time from the standard case.
The table shows very little sensitivity of the probabilities to
TABLE
7:
Interp retation of Tabl e
Probability of Answering Inflation:
Date of
Poll
6
•
Independent Variable
Regression
INCOME
EDUCATION
//
10-15,000
15 -20,000
20-25,000
13-15
12
1
16+
75
1
(.382)
(.393)
(.404)
.393
.426
.396
Nov. 75
2
(.533)
(.540)
(.547)
.540
.587
.540
Dec.
77
3
(.621)
(.629)
(.637)
(.629)
(.648)
(.602)
Feb.
78
4
(.655)
(.659)
(.663)
(.659)
(.680)
(.677)
May
78
5
(.744)
(.752)
(.760)
.752
.799
.767
Feb.
18-20
35-44
SEX
B/H
AGE
Black/Hispanic
65+
White
Male
Female
Feb.
75
1
(.435)
(.393)
(.449)
(.343)
ijov.
75
2
(.509)
(.540)
(.571)
.396
.54^
:.5':0)
(.517)
Dec.
77
3
(.568)
(.629)
(.634)
.378
.629
:.629)
(.591)
Feb.
7S
4
.538
.659
.594
.421
.659
(.659)
(.627)
May
78
5
.626
.752
.723
.476
.752
(.752)
(.712)
(.334)
(.393) :.393)
EXPECTED INFLATION
EXPECTED UN EMPLOYMENTr
--
Less
Same
More
0%
5%
1
10%
75
1
(.429)
(.393)
(.357)
.329
.393
.457
Nov. 75
2
.595
.540
.445
.473
.540
.607
Dec. 77
3
.711
.629
.547
.540
.629
.718
Feb.
78
4
.760
.659
.558
.613
.659
.705
May
78
5
.799
.752
.705
.720
.752
.784
Feb.
Note:
Pa:rentheses
tilae
period.
inc iicate tllat
tlle
V ariables
c
lid not:
1
ent er sigilificantly
ii1
that
1
i
..,1,
-16-
income and sex.
A doubling of income from $12,500-$25,000 would
increase the predicted probability by less than 2% in most cases.
The average difference between males and females is under 4%.
effect of education is about the same.
The
The average difference between
high school graduates and those who receive some college education but do
not graduate is about 3%, with additional education erasing this difference.
The effect of age on attitudes to the social cost of inflation
and unemployment can be substantial at times and seems to be growing.
By the end of 1978 there was an estimated 13% higher probability
that someone age 35-44 would see inflation as the more serious problem
than would someone 18-20.
The parabolic nature of the age relationship found
in the later periods shows middle aged persons to be the most concerned about
inflation.
B/H has the strongest quantitative effect on the estimated
probabilities.
With the exception of the first peridd, whites have at least
a 15% higher predicted probability of choosing inflation as the more serious
By late 1978 the difference had reached 28%.
problem than do blacks and hispanics.
The list two variables in Table
7,
the expectional variables, are
also seen to have large effects on the inf lation-unemplojnnent response.
Increasing inflation expectations from 0% to 10% increases the
probability of answering inflation by an average of 12%, but this effect
is larger in the earlier polls. ..The effects of differences in
expected unemployment are of the same magnitude.
Those who expect
unemployment to fall have on average about a 14% higher probability of
naming inflation as the more serious problem than do those who expect
-17-
linemployment to rise.
In summary then, the regressions show that only the expectational
variables and B/H contribute consistently, significantly,
?.ud
with
large empirical effects to determining individuals' perceptions
of the social costs of unemployment relative to inflation.
19
Characteristics which might have been expected to shape attitudes such as income, education, sex - have an empirically small and generally
insignificant effect.
Age is the only demographic variable besides
B/H that has a quantitatively large effect, but it does so in only
two out of five periods.
In general the pattern of responses is
consistent with the assumption that individuals are answering the polls
in terms of a view of society's,
IV.
rather than their own, interests.
Changes in Poll Results Over Time
.
The cross sectional results of the previous section show that
expectational variables and ethnic background are the most important
variables determining poll responses at a moment of time.
In this
section we examine the effects of the past history of inflation and
unemployment, along with expectational variables, on inflation aversion.
The dependent variable in the regressions reported below is the
percentage of respondents who state that inflation is the more serious
problem
— this
is the aggregate index of inflation aversion.
time series behavior is shown in Figure
rates of inflation and unemployment.
3,
Its
along with the actual
The independent variables are
expected inflation, expected unemployment and lagged values of inflation
and unemployment.
The expected inflation variable is the average
Figure
3
.
13.©
10.8
Notes
;
The inflation index is the percentage responding "Inflation" plus
half of the percentage responding "Both".
The inflation rate is for the previous six months, at an annual rate
The unemployment rate is for the previous month
-18-
expectation based on responses to the question used in the cross section.
We have constructed a quantitative expected unemployment variable from
the qualitative question used in the cross section.
Appendix
2
describes
the method of construction of the expected unemployment variable and
presents both expectat ional series over the entire period for which they
are available.
The results of the regression described above are
(1)
PI = 101.5 + 1.38 DPE + .61 DP + .15 DP
(16.1)
(.63)
(.37)
- .01DPE_
(.43)
(5.20)
R^ = .964 SER = 2.29
(1.61)
SSR = 42.022
- .68 DPE
(.67)
-3.18 UE - 9.46 U + 4.51 U_^ + .82
(3.60)
21
(.54)
\]_^
(2.1)
D.W. = 1.84
number of observations = 18
PI is the percentage resonding inflation, where those answering
both were divided equally among inflation and unemployment.
DPE is the expected inflation rate, over the next year, expressed
as a percentage
DP is the inflation rate since the last quarter, at an annual rate
(for the February poll it is the inflation rate from October to
January, etc.)
UE is the expected unemployment rate for the next year, expressed
as a percentage.
U is the actual unemployment rate in the month preceding the poll
The subscript -1 indicates a one quarter lag. Standard errors are
in parentheses.
-19-
As expected, an increase in the expected rate of inflation increases
inflation aversion while an increase in expected unemployment lowers it.
Equation
month
1
also shows that an increase in the level of unemployment in the
preceding the poll lowers inflation aversion.
In interpreting the
coefficient on the lagged unemployment terms, note that -9.46 U + 4.51 U
^
is the same as -4.95 U - 4.51(U-U ,),
so that the coefficients can be
interpreted as showing that the change as well as the level of unemployment
affect inflation aversion. The twice lagged unemployment rate is in-
Turning to the inflation variables, it is clear that higher
significant.
Inflation in the past, given expectations, increases the percentage
The similar magnitudes of the coefficients on DP
responding inflation.
and DPE
and the apparent insignificance of DP
^
and DPE
^
suggest
that only unanticipated inflation over the two quarters previous to the
poll affect responses.
Equation
the
2
below presents the results of re-estimating equation
imposing
1
constraints that the coefficient on the twice lagged unemployment rate
is zero and that lagged inflation matters only if it was unanticipated.
(2)
PI = 106.1 + 1.42 DPE + .80(D2P - DPE
(9.1)
(.51)
R^ = .974 SER
=1.95
(.27)
SSR = 45.622
)
- 5.09 UE - 2. 34 U
(2.06)
D.W.
(1.94)
--
4.36 (U-U
(1.86)
=1.85
number of observations = 18
D2P is the rate of inflation over the last six months, expressed
as a percentage, at an annual rate.
Figure
4
presents the actual level of the dependent variable and that
predicted by equation
2.
.)
FIGURE
A:
Predicted and Actual Levels of Inflation Aversion; 1974-1979
(Square = Actual; Diamond = Predicted, from equation
(2))
78.
.
1
y^
r^
L
•
66
.
•
•
54
i^r
-^^•^fc^i^^^^^^
»r'
"
'S,
=3^
y^
v/
/ n/
V
=/
p-^r
1
f
42
:
30.
/
/
*v/
w
3
1974
1975
1976
3
1977
1978
3
1979
-20-
V.
What Do the Poll Results Mean ?
There are at least two types of information that economists would
•like to get from public opinion surveys.
The first arises from the
profession's difficulties in quantifying the costs of inflation: the
polls appear to show that the costs of inflation are high relative to
unemployment.
Since rational policy making requires estimates of the
costs of inflation and unemployment on alternative policy paths,
evidence on their relative costs would be of value in choosing optimal policy.
Second, on the positive side, election studies showing that some
opinion poll results are reflected in voting behavior,
media attention
to the polls, and political rhetoric all suggest that the polls affect
economic policy.
For that reason, knowledge of the determinants of poll
responses should increase understanding of economic policies that are
likely to be accepted, as well as their subsequent political effects.
The evidence reviewed in this paper suggests that the poll results
are of little help in providing the first type of Information.
They
do Indicate that there is a significant difference in people's perception
of the social and private costs of inflation relative to unemployment,
that the perceived relative social costs do depend (in the expected way) on both
the actual and expected amounts of inflation and unemployment, and that
at a given point in time there is a remarkable amount of agreement among
individuals about the relative social costs of inflation and unemployment.
However, due to the lack of a well specified trade-off to be considered
by the respondents, this evidence on the relative costs of inflation and
-21-
unemployment does not enable one to make any statements about the
desirability
of certain policy alternatives.
This is most forcefully
brought out by the apparently inconsistent result that inflation is
generally supposed to be the more serious problem while
it
is preferred
that policy be directed at reducing unemployment.
Two additional points should also be made about the evidence from the polls with
regard to the costs of inflation.
First, there are survey questions which ask about
the desired direction of policy in addition to asking about which is a
more serious problem.
not
The results are clearly that the public is in general
willing to fight inflation at the cost of higher unemployment - and that is
exactly
the choice that has been offered by most economists and governments.
Unfortunately, it is not possible to discover the estimates of social
costs that might
underlie this response.
As pointed out in Section II,
the questions dealing with the welfare costs are sufficiently vague to
allow a rational person to answer that inflation is a more serious problem
while feeling that policy should be directed at lowering unemployment.
Second, the failure of the survey evidence to provide useful evidence
on the social costs of inflation does not depend on the poll responses
being either unstable or sensitive to the wording of the questions.
The
evidence reviewed in this paper, at least with respect to the questions
about whether inflation or unemployment is a more serious problem suggests
that responses are not very sensitive to the precise wording of the questions.
'Jot
did we find much evidence that supports the contention that people
fail to understand the effect of inflation on nominal incomes.
Turning to the second type of information economists may hope to get
from public opinion polls, i.e. what determines the poll responses, the
..
.
.„
.....
-22-
strong conclusion is that the responses are very well explained by inflation
and unemployment themselves.
it is not
only actual,
affect poll responses.
Ijut
The main point of interest here is that
also expected, inflation and unemployment that
To reduce concern over inflation it is necessary
to reduce expected inflation and/or lower the amount of unexpected inflation.
Of the two, expectations for the future appear to have the more significant
effect.
To lower concern over unemployment it is necessary to lower expected
unemployment, the current level, and/or the current rate of change of
the unemployment rate.
-23Appendix
1
TABLE Al
Gallup Poll:
"What do you think is the most important problem facing this
country today?"
(P'ercent
Date
citing)
Inf la tion
(Percent citing)
Unemp loyment
April 1939
36
November
24
June- July 1946
46
January 1947
13
July
24
September
37
March 1948
2
8
June
23
September 1949
11
12
9
6
March 1950
15
10
September 1951
24
March 1954
13
16
September 1956
10
4
October
13
3
May 195/
21
Augus t- S ep t emb er
22
November
June 1955
September 1958
11
February 1959
17
9
April
15
10
September
13
3
.,-24-
Date
Inflation
Unemployment
February 1960
8
5
June
2
25
."962
10
8
March 1963
4
11
September
3
5
April 1964
5
9
June
3
6
Augus t
3
4
September
6
4
March 1965
3
2
May
4
3
September
5
October-November
6
November
3
May 1966
16
August
16
October-November 1967
16
April
May 1968
8
June- July
9
August
7
January 1969
9
2
May 1970
10
July 1972
16
7
May 1973
39
4
September
50
2
-25-
Date
Inflation
Unemployment
January 197A
25
5
May- June
48
2
August
77
2
September
81
2
October
79
3
February-March 1975
60
20
July
51
21
October
57
21
January 1976
47
23
April
38
24
October (early)
45
33
October (late)
47
31
March 1977
58
39
July
32
17
October
35
25
February 1978
33
17
April
54
18
July
60
14
September
59
12
October
75
8
-26-
TABLE A2
Michigan Poll:
"Which of the two problems - inflation or unemployment - do
you think will cause the most serious economic hardship for
people during the next year or so?"
Date
(Percent citing)
(Percent citing
Inflation
Unemployment
(Percent citing
Both
1974: III
67.0
25.1
5
1975: I
29.7
60.8
6.2
:
II
28.8
64.1
4.7
:
III
43.7
38.6
16.5
:
IV
45.6
36.1
15.8
44.5
36.7
16.8
1976: I
:
II
46.2
36.5
13.5
:
III
47.6
38.9
10.7
:
IV
48.6
38.1
9.7
54.2
35.0
7.9
1977: I
:
II
61.1
30.4
6.2
:
III
56
32
7
:
IV
52
37
9
1978: 1
55
33
9
2
62
26
8
3
56
33
8
4
64
27
7
5
63
26
8
6
69
22
7
7
63
25
8
8
66
24
8
9
65
26
5
10
73
18
6
11
65
23
8
67
24
7
1979:2
-27-
Appendlx
2
In order to construct the expected unemployment series we assume that at
time
t,
a respondent has a subjective probability destribution for the unemploy-
ment rate in each of the four upcoTiing quarters.
tions for individual
U
,
,
at time
i
are denoted by U
t
The means of these distribu-
^,
U
,
U
and
_
It is assumed that a respondent will answer that he expects the
.
unemployment rate to rise if UE^ = 1/4 (U^
t
where U
+U^
t+1
+
+
L'^
t+2
t+3
U^
U^(l+d)
>
)
t^
t+4
is the current unemployment rate and d is a parameter to be estimated.
2
Likewise, a respondent is assumed to answer that he expects the unemployment
rate to fall if UE^
(1-d) and to answer that he expects no change in
U
<
the unemployment rate if U (1-d)
_<
We also assume that the UE
<_
U (1+d).
are distributed according to a triangular
distribution in each time period.
by equation
UE
The probability density function is given
below.
(1)
f^(x) =
(1)
A4/(b^. - a^)2
(^(4/(b^ -
£
where
a
< b
a^
<
(x - a^.)
if
a^ x
(b^ -x)
if
^t
'^
-L-—L
<
^t
<
X
<
b^.
1
This distribution is symmetric about its mean, UE
=
a
t
+ b
t
,
and has
2
variance
S
2
= (b
- a
2
)
/24.
Using these assumptions, it is not hard to show that in a random sample
the expected value of the percentage answering that they expect a rise in the
unemployment rate is 1-F
(l!|
(1+d)) where F
(x)
is the
cujnulative distribution
function for a random variable whose density function is given by (1).
Similarly
the expected value of the percentage expecting a fall is F (U (1-d)). If we denote
the expected value of the percentage expecting a rise and the percentage expecting
aTaH
aTai:
C3),
in period
t
as P
and P
respectively, we arrive at equations (2) and
-28-
^t ^"t ^^ + ^^
(2)
Pit "
(3)
P2, = F^ (U^ (1 - d)
By setting
"
)
equal to the percentages actually observed in
and P
P,
^
'^'
'
period
t,
"
equations (2) and (3) become two equations with two unknowns (a
and b
Furthermore, the equations can be rewritten so that they are linear in UE
and S
.
To see this note that the right hand side of equation (2) is the
probability that a random variable
(1), will be greater than U
a random variable Y =
(1
(X - UE
X>
+ d).
)
/S
with density function given by equation
This is equal to the probability that
is greater than [U
(1
+ d) - UE
Since the probability density function for Y is independent of a
we can rewrite equation
{2)\
Uj.
(1
+
(2)
d)
/S
]
and b
.
,
as
- UE^ = G
(1 - P^^)
where G is the inverse
S^
cumulative distribution function
for Y.
Similarly, equation (3) can be rewritten as
(3')
Substituting
Uj.
S
(A)
(1 - d) - UE^ = G(P2j.)
Sj.
out of equations (2') and (3') yields
UE^ = U^ [1 + d (G(P2t) + G(l - '^i^))f(^G(^2t^ - G (1
-
P^.^))]
In order to use equation (4) to construct an expected unemplo3nnent
variable one more piece of information is needed; d is still an unknown.
We estimated d by imposing rationality on the sample for the time period
1960 to 1978.
U
.„
t+2
+ U
_
t+3
+U
That is, we suppose U
,)
t+4
and V
-'TT^ = d ir^
+ V
where U
= 1/A (U
,+
is a random variable with mean zero that is
t
uncorrelated with both U
"Uj.
= UE
and UE
.
This implies a regression model
(GCP^^) + G(l-Pj^j.))/(G(P2^) - G(l-P^p) + V^.
3
)
-29-
The results of estimating this regression are presented in equation (5) below.
u^ - U^ = .116
(5)
Uj.
(G(P2j.)
+ G(l
-
Pit^)/(G(P2t> " Gd-P^^.))
(.02)
SER = .61
Standard errors in parentheses
R^ = .38
D.W.
=
0.70
The fact that d is estimated to be .116 in our model implies that if
the unemployment rate is 5%, respondents answer that they expect no change when
UE
is between A. 4% and 5.6%.
This range is not totally unreaisonable.
Durbin -Watson statistic is to be expected since the error
low
(U
The
-
UE
)
contains four different expectational errors; one for each of the four upcoming
quarters.
Table A3 contains the expected unemployment variable conctructei^ according
to equation (4) with d= .116.
Table A4 contains the expected inflation variable.
The procedure followed here is similar to that of Carlson and Parkin
in "Inflation Expectations", Economica 42, pp. 123-138.
2
The survey question is stated, "How about people out of work during the
coming 12 months - Do you think that there will be more unemployment than now,
about the same, or less?"
3
Since the survey includes close to 1500 respondents, the observed
percentages stiouid be
very close to their expected values, i.e., the
true population percentages.
4
The random variable Y has an inverse cumulative destribution function
of the form
G(Y) =
( - (6)^^^ + (12Y)
(6)-^^^ -
-""^^
(12(1-Y))-'-''^
0< Y <
1/2 <
1/2
Y < 1
For a thorough explanation of how this variable was constructed see Juster,
"A Note on the Measurement of Price Expectations", 1978
manuscript, University of Michigan.
F.T. and Comment, R.
,
-30Table A3
:
Expected Unemployment
1st Qtr
2nd Qtr
.
3rd Qtr
.
4th Qtr
1960
NA
NA
NA
5.9
61
NA
6.4
NA
5.9
62
NA
5.4
5.7
5.6
63
5.9
5.9
5.4
5.7
64
5.6
5.1
NA
NA
65
5.0
NA
4.1
3.9
66
3.4
3.7
3.7
3.6
67
3.8
NA
3.7
NA
68
NA
3.5
3.5
NA
69
3.5
3.5
3.7
4.0
1970
5.0
5.5
5.7
6.8
71
6.4
6.5
6.4
6.3
72
5.7
5.7
5.6
5.5
73
5.6
NA
5.1
5.1
74
6.5
5.6
6.5
8.0
75
9.4
9.4
8.3
8.5
76
7.4
7.2
7.5
7.5
77
6.9
6.6-"
6.9
6.9
78
6.2
6.2
6.2
6.1
79
6.0
Note:
NA indicates not available.
-31-
TABLE A4
Date
:
Expected Inflation
1st Quarter
2nd Quarter
3rd Quarter
Ath Quarter
19A8
1,27
-.33
1.01
.15
1949
-.70
-.62
-.5A
-.43
1950
-.33
.39
1.11
1.83
1951
2.57
1.07
1.50
1.94
1952
1.62
.64
.16
-.32
1953
-.30
-.28
-.26
-.23
1954
-.25
-.28
-.26
-.24
1955
-.02
.20
.52
.84
1956
.89
.95
1.27
1.36
1957
1.28
1.21
1.17
1.13
1958
.75
.38
.79
1.21
1959
1.37
1.53
1.66
1.79
1960
1.67
1.56
1.40
1.24
1961
1.88
1.88
1.91
1.94
1962
2.25
1.94
1.73
1.91
1963
2.54
2.40
2.27
2.14
1964
2.46
2.48
2.50
2.52
1965
2.54
2.61
2.69
2.72
Table A4 continued...
Date
-32-
1st Quarter
2nd Quarter
3rd Quarter
4th Quarter
1966
2.63
3.86
4.78
4.34
1967
3.91
4.28
4.65
5.09
1968
4.78
4.47
5.11
3.87
1969
4.53
5.48
5.49
4.07
1970
5.51
6.04
4.32
5.40
1971
5.21
4.65
3.53
4.01
1972
4.18
3.04
3.87
3.91
1973
6.78
6.11
7.63
6.43
1974
9.97
9.41
9.26
10.66
1975
5.40
3.44
7.81
7.24
1976
5.80
5.82
5.40
4.80
1977
7.40
6.90
7.20
6.90
1978
7.80
8.20
9.80
8.90
1979
10.70
-33-
A ppendlx
3
The statistical model that underlies regressions 1-5 is an ordered
That is, individual
probit model.
i
is assumed to have a certain
marginal rate of substitution betveen inflation and unemploymeit
.
This
is assumed to be linear in the parameter vector B so that MRS, = X. B + e.
where X
is a 1 x n vector of pe-:fonal cliaracteristics and e.
distributed random variable with mean zero and variance
assumed the e. and
e.
are uncorrelated for
i
2
S
.
not equal to j.
is a normally
It
is
Given this
framework, an individual is presumed to answer that inflation is a more
serious problem than unemployment if MRS
t_.
If t,
>
.
t„ for some unknown parameter
MRS. < t„ (for unknown parameter t.) the respondent is
<
assumed to answer that both inflation and unemployment are serious problems.
If MRS
,
< t.
,
the respondent is assumed to answer that unemployment is the
more serious problem.
As in the regular probit case, writing out the
likelihood function shows that only B/(S
have normalized
2
S
h
2
)
is identified.
Hence, we
= 1.
A complication that arises in the ordered probit case that is absent
In the regular probit case is that
if we include a constant
and we denote its coefficients
B,
all identified.
,
and the threshold points are not
Specifically one of the coefficients must be normalized.
We have chosen to normalize
estimates of B^ -
B.
term in X.,
t.
t^
=
and t- ~ t,.
so that we present
(in Table A5)
This latter term is thus the range of
indifference between inflation and unemployment.
-3A-
Appendlx
4
The variables used in the regressions reported in Tables
scaled in the following manner:
B/H:
Sex:
Education:
Education +:
Age:
1
-
White
2
-
Black/Hispanic
1
-
Male
2
-
Female
1
-
0-5 years
2
-
6-11
3
-
12
4
-
13-15
"
5
-
16+
"
1
- 16+
"
^
otherwise
1
-
18-20 years
2
-
21-24
3
-
25-34
4
-
34-44
"
5
-
45-54
"
6
-
55-64
"
7
^
65+
"
••
6
and
7
were
-351
-
18-20 years
4
-
21-24
"
9
-
25-34
"
16
-
35-44
"
25
-
45-54
"
36
-
55-64
"
49
^
65+
Age
Expected Unemployment:
1 -
less
3 -
same
5 -
more
Income:
1
-
$0 -4,999
(1975)
2
-
$5,000 - 9,999
3
-
$10,000 - 14,999
4
-
$15,000 - 19,999
5
-
$20,000 - 24,999
6
-
$25,000 - 29,999
7
-
$30,000 - 34,999
9
-
$35,000 +
:
-36-
Income
$0-4,999
(1977-78)
$5,000-9,999
$10,000 - 14,999
$15,000 - 19,999
$20,000 - 24,999
$25,000 +
Expected Inflation
< 0%
0%
1-4%
5%
6-9%
10 +%
The income variable appears in two forms since the responses are tabulated
differently at different times by the Michigan Survey.
for the total family unit and is before tax.
The income variable is
All other variables are self
explanatory.
Table A5 contains the estimates of parameters for regressions 1-5
not presented in Table
6
in the text.
which_are
-37TABLE A5
:
Estimation Results from Regressions 1-
Regression
Date
Constan t
-.33
5.
Threshold Point
Feb.
75
1
Nov.
75
2
.46 (1.4)
.39 (13.9)
Dec.
77
3
.76 (1.9)
.21 (8.9)
Feb.
78
4
.69 (1.7)
.17
5
.68
.18 (7.9)
May 78
(0.7)
(1.8)
.15 (6.4)
(7.6)
A
,
REFERENCES
S.
Fischer and
F.
Modigliani, "Aspects of the Costs of Inflation",
presented at a conference on The
Economic Crisis of the 1970' s:
Lessons for Stabilization Policy. Baden, Austria, Sept. 1977
B.S.
Frey and
F.
Schneider, "An Empirical Study of Politico-Economic
Interactions in the United States", Rev. Econ. Statist ., May 1978,
174-183.
G.H. Gallup, The Gallup Poll
,
New York, Vol 1,2,3, 1972.
"Economic Interest and the Politics of Macroeconomic
D.A. Hibbs,
Policy", Center for International Studies, M.I.T., unpub.
,
1976.
"The Mass Public and Macroeconomic Performance: The
,
Dynamics of Public Opinion Toward Unemployment and Inflation",
Amer. J. Polit.
J.
Sci.
November 1979, 23, 705-731.
,
Huizinga, "Testing the Rationality of Inflation Forecasts:
Theory and Evidence from the Survey of Consumer Attitudes"
unpublished, December 1979.
T.
Johnson, "Qualitative and Limited Dependent Variables in Economic
Relationships", Econometrica
,
May 1972, 455-462.
F.T. Juster, "Why is Inflation Bad?", Econ.
1978-79,
b,
1,
Outlook U.S.A
.
,
Winter
16-18.
and P.Wachtel, "A Note on Inflation and the Saving Rate",
Brookings Papers on Econ. Act
G.
.,
1972,
3_,
765-778.
Katona, "Why Do We Have Inflation at Present", Econ. Outlook U. S.
Winter 1978-79,
J^,
.
18-19.
"The Psychology of inflation," in Richard T. Curtin, ed.
,
Surveys of Consumers, 1974-75 , Survey Research Center, Michigan.
S.M. Lipset, "The Wavering Polls", Pub.
Interest , Spring 1976, 43, 70-89.
M.C. Lovell, "Why Was the Consumer Feeling So Sad?", Brookings Papers on
Ecpn. Act,
,
1975,
2^,
473-479,
FOOTNOTES
1.
This question has been asked oy the University of Michigan Institute
for Social Research since 1974.
We shall henceforth describe this as the
standard question,
2.
The Roper Center at the University of ConnecticuL collects opinion poll
results systematically, and makes the results available in a convenient form.
The journal Public Opinion reports interesting findings from polls on a
regular basis.
3.
Appendix
to 1978.
1
contains the dates and responses from Gallup polls from 1939
They are reproduced
number 160, page
19.
from the Gallup Opinion Index (November 1978),
Douglas A. Hibbs (1979)
;.
.
.
has also examined this data.
4.
However, given the presence of other problems than inflation and
unemployment as possible answers to the Gallup question, it is not certain
that the responses shown in Figure
1
would be obtained if individuals were
asked merely to rank the seriousness of inflation and unemployment.
5.
This statement is based on an econometric examination of the complete time series
for the periods in which they overlap.
6.
Results similar to those of the Harris poll can be found in the New York
In April 1978 the question was asked "In the last few
Times/CBS poll.
years would you say that inflation or unemployment has caused a more
serious hardship for your family?"
unemployment.
The responses were 92.6% inflation and 7.4%
Although no question about which was a more serious problem
for the country was asked at the same time, the April 1977 poll did include
such a question.
7.
The
answers
th-^n
were 59.5% inflation and 40.5% unemployment.
For evidence on how well the public understands the actual relationship
between inflation and unemployment, see Iluizinga (1979).
FOOTNOTES (2)
8.
Answers to the policy choice question could also be made on the
basis of the Intersection points of the PP curve with the axes.
If
Indifference curves are parallel and linear, and the PP curve is linear,
then the policy choice based on relative slopes at A is the same as the
policy choice based on intersections with the axes.
9.
See George Katona (1975).
10.
See Sejrmour Martin Lipset (1976) for a discussion of this phenomenon.
11.
See for example Gallup Opinion Index (November 1978), Report 160,
PP 13-20.
12.
However, since the Michigan survey does not follow individuals through time,
we are unable to do a combined cross-section time-series analysis.
13. The other polls are omitted to keep the table simpler.
We will footnote
circumstances in which the results in the remaining two polls differ
substantially from those presented in the table.
Regression results are
presented below for all five polls.
14.
A longer version of Table
5
is available on request.
The present version
either contains data to which we want to refer later or characteristics that
are not included in the regression of Table 5, but which nonetheless are of
interest.
15. If people
were answering in their personal interest we would expect the
unemployed to be considerably more worried about unemployment.
A possible
exception to this would be if people thought inflation caused unemployment.
This seems to be contrary to the evidence of the previous section, however,
which showed people to perceive a trade-off between inflation and unemployment.
16. The sole exception is May 1978.
.
FOOTNOTES (3)
17.
The expected inflation variable is the response to a question about
inflation expected for the next year asked of all respondents who are asked
the inf lation/uneraplojraient question.
The expected unemployment variable is
from a question which asks if the respondent expects more, less, or the
same amount of unemployment in the upcoming year.
18.
The model is the same as described by Thomas Johnson (1972).
19.
The powerful role of expectations in the regressions raises the question
of whether expectations reported to the pollsters might merely represent
rationalizations of underlying attitudes rather than true expectations.
There
are a number of ways in which the notion of rationalizing expectations
might be formalized; the method we have chosen suggests that expectations
are not obviously rationalizations.
20.
Details are available on request.
Since the racial make-up of the sample does not change much over the
sample, omitting a measure of racial composition should not affect the time
series results.
21
This regression was run after earlier experimentation showed that annual
rates of Inflation and unemployment did a poor job of explaining the behavior
of the inflation aversion index.
It was
deceided to use lags of up to two
quarters in light of the limited number of observations we have.
Similarly,
we did not experiment with the length of lag on the expected unemployment
variable because we find the argument that lagged unexpected unemployment
rather than lagged actual unemployment should affect inflation aversion much
weaker than that relating to unanticipated and actual Inflation.
However,
we h^ve run regressions including lagged unanticipated unemplo^-ment in place
of lagged actual unemployment and find that these perform about as well
as the regression reported in the text.
FOOTNOTES (4)
22.
Examination of the sums of squared residuals in equations
that the F-statistic has a value of .23 compared with F
1
and
2
shows
(4,8) = 3.84,
the critical value using a 5% significance level.
23.
Bruno Frey and Friederich Schneider (1978) examine the effects of economic
conditions on presidential popularity as measured by opinion polls, and
the effects of popularity on policy.
Date Due
JiSU
3
TD6D
DM
mS
ASM
3
TDflD DOM
MIS
flbE
D
Zi*!
TDSD DO M
3
b
M7fi
MiT LIBRARIES
Si/"^
3
ms
TDfiD DDM
fi7
MIT LIBRARIES
31-i
3
TDflD DDM
MIS
flflfi
Ol^V
3
TD6 D
D
DM MIS
6"it,
MIT LISfiAHlES
3(4^
3
TDflD DDM
MIS
"iDM